REPORTS

8. H. Lammer et al., Planet. Space Sci. 54, 1445 micro-ropes, which appear not to have a magnetic 21. H. Y. Wei, C. T. Russell, T. L. Zhang, M. K. Dougherty, (2006). reconnection source (21). Icarus 206, 174 (2010). 9. S. Barabash et al., Nature 450, 650 (2007). 13. T. L. Zhang et al., Nature 450, 654 (2007). 22. T. L. Zhang et al., Planet. Space Sci. 54, 1336 10. The bipolar signature in our rotated coordinate system 14. C. T. Russell, O. Vaisberg, in Venus,D.M.Hunten,L.Colin, (2006). is in the BY component here instead of the BZ component, T. M. Donahue, V. I. Moroz, Eds. (Univ. of Arizona Press, 23. S. Barabash et al., Planet. Space Sci. 55, 1772 as it would be in the usual coordinate system used for Tucson, AZ, 1983), pp. 873–940. (2007). a terrestrial magnetotail plasmoid (18). 15. J. L. Phillips, D. L. McComas, Space Sci. Rev. 55, 11. T. L. Zhang et al., Geophys. Res. Lett. 37, L14202 1 (1991). Acknowledgments: The work in China was supported by (2010). 16. A. Fedorov et al., J. Geophys. Res. 116, (A7), A07220 National Natural Science Foundation of China grants 12. An earlier study found a flux rope in the Venus tail, (2011). 41121003 and 41174156. The work at Graz was supported but reconnection was rejected as a possible source 17. V. Angelopoulos et al., Science 321, 931 (2008). by Austrian Science Fund (FWF: I429-N16). The work at UCLA because of the inconsistent orientation of the rope 18. M. B. Moldwin, W. J. Hughes, J. Geophys. Res. 96, (A8), was supported by NASA under research grant NNX10AV29G. axis (19). In our case, the axis is consistent with 14051 (1991). reconnection. Flux ropes that have been seen on the 19. J. A. Slavin et al., Geophys. Res. Lett. 36, L09106 23 November 2011; accepted 8 March 2012 nightside of Venus are much larger than those in the (2009). Published online 5 April 2012; dayside ionosphere (20) and cannot explain these 20. C. T. Russell, R. C. Elphic, Nature 279, 616 (1979). 10.1126/science.1217013

Chester Lake and all the rocks near Greeley Ancient Impact and Aqueous Haven have similar textures. They are brecciated, with dark, relatively smooth angular clasts up to Processes at Endeavour Crater, Mars ~10 cm in size embedded in a brighter, fractured, fine-grained matrix. Some outcrops, notably Chester Lake, show fine-scale lineations in the S. W. Squyres,1* R. E. Arvidson,2 J. F. Bell III,3 F. Calef III,4 B. C. Clark,5 B. A. Cohen,6 7 8 5 9 10 11 matrix and alignment of some clasts (Fig. 2). L. A. Crumpler, P. A. de Souza Jr., W. H. Farrand, R. Gellert, J. Grant, K. E. Herkenhoff, Pancam spectra of the matrix exhibit a gradual J. A. Hurowitz,4 J. R. Johnson,12 B. L. Jolliff,2 A. H. Knoll,13 R. Li,14 S. M. McLennan,15 16 16 4 17 1 3 decrease in reflectance toward 1000 nm. The clasts D. W. Ming, D. W. Mittlefehldt, T. J. Parker, G. Paulsen, M. S. Rice, S. W. Ruff, can show specular reflections, have a relative- C. Schröder,18 A. S. Yen,4 K. Zacny17

ly deep absorption at 934 nm, and have a shallower on June 1, 2012 535-nm absorption than the matrix materials, The rover Opportunity has investigated the rim of Endeavour Crater, a large ancient impact consistent with relatively unoxidized basaltic ma- crater on Mars. Basaltic breccias produced by the impact form the rim deposits, with stratigraphy terial containing low-Ca pyroxene. similar to that observed at similar-sized craters on Earth. Highly localized zinc enrichments in The matrix of Chester Lake is easily abraded. some breccia materials suggest hydrothermal alteration of rim deposits. Gypsum-rich veins cut A portion of Chester Lake dominated by matrix sedimentary rocks adjacent to the crater rim. The gypsum was precipitated from low-temperature was abraded to a depth of ~2.5 mm with the aqueous fluids flowing upward from the ancient materials of the rim, leading temporarily to rover’s Rock Abrasion Tool (RAT). Resistance to potentially habitable conditions and providing some of the waters involved in formation of abrasion is quantified using specific grind energy, the ubiquitous sulfate-rich sandstones of the Meridiani region. the energy required to abrade away a unit volume www.sciencemag.org of rock. The specific grind energy for Chester fter more than 7 years in operation and the sulfate-rich sedimentary rocks explored by Lake was ~1.5 J mm−3. Representative values for 33 km of traversing, the Mars Explora- Opportunity for most of its mission (2, 3). En- weak terrestrial materials are 0.7 to 0.9 J mm−3 A tion Rover Opportunity has reached En- deavour was chosen as a target because the rocks for chalk and 4.8 to 5.3 J mm−3 for gypsum (10). deavour Crater. Endeavour is ~22 km in diameter there record an ancient epoch in martian history, Chester Lake is substantially weaker than all but andformedinNoachian(1) materials that predate and because orbital infrared data show that phyl- 1 of the 14 diverse rocks abraded by Spirit at losilicate minerals are present in portions of the Gusev Crater (11) but is comparable to the

crater rim (4). sulfate-rich sandstones at Opportunity’s landing Downloaded from 1Department of Astronomy, Cornell University, Ithaca, NY 12 2 Opportunity arrived at Endeavour Crater on site ( ). 14853, USA. Department of Earth and Planetary Sciences, 5 Washington University, St. Louis, MO 63031, USA. 3School of sol 2681 ( ) of its mission, at a low-lying seg- At Chester Lake, we used the Alpha Particle Earth and Space Exploration, Arizona State University, Tempe, ment of the rim, ~700 m in length, named Cape X-Ray Spectrometer (APXS) to measure the AZ 85287, USA. 4Jet Propulsion Laboratory, California Institute York (Fig. 1). Shoemaker Ridge (6)formsthe elemental composition of both the matrix (after 5 of Technology, Pasadena, CA 91109, USA. Space Science In- spine of Cape York and is the type locality for abrasion by the RAT) and one of the clasts. Mea- stitute, Boulder, CO 80301, USA. 6NASA Marshall Space Flight Center, Huntsville, AL 35812, USA. 7New Mexico Museum of the Noachian materials of the rim, which we call surements were also made of three targets near Natural History & Science, Albuquerque, NM 87104, USA. 8Hu- the Shoemaker formation. Opportunity first ar- Greeley Haven: Transvaal and Boesmanskop man Interface Technology Laboratory, University of Tasmania, rived at Spirit Point, the southern tip of Cape (both matrix) and Komati (a clast). All are similar Launceston TAS 7250, Australia. 9Department of Physics, 10 York, and then traversed northward 851 m before to one another in composition, and all are similar University of Guelph, Guelph, Ontario N1G 2W1, Canada. Cen- 7 ter for Earth and Planetary Studies, Smithsonian Institution, stopping at Greeley Haven ( ) at the northern to the basaltic sand typical of the Meridiani re- Washington, DC 20560, USA. 11U.S. Geological Survey, Astro- end of Cape York to spend the martian winter. gion (Table 1). The major elements (Na, Mg, Al, geology Science Center, Flagstaff, AZ 86001, USA. 12Applied Instruments of Opportunity’s Athena payload Si, Ca, and Fe) are mostly within 10 weight per- Physics Laboratory, Johns Hopkins University, Laurel, MD 20723, 8 9 13 ( , ) were used to investigate materials within cent (wt %) of the basaltic sand composition, and USA. Botanical Museum, Harvard University, Cambridge, MA the Shoemaker formation, including the bedrock all but a few are within 20 wt %. 02138, USA. 14Department of Civil, Environmental, and Geodetic Engineering, Ohio State University, Columbus, OH 43210, USA. outcrop Chester Lake (Fig. 2) near the southern Fe/Mn ratios of the matrix range from 40 to 15Department of Geosciences, State University of New York, Stony end of Shoemaker Ridge, and several bedrock 44, and “Mg numbers” [100 × molar Mg/(Mg + Fe)] Brook, NY 11794, USA. 16Astromaterials Research and Explora- targets near Greeley Haven at the northern end. of all samples range from 41 to 48. These are tion Science (ARES), NASA Johnson Space Center, Houston, TX Although separated by more than half a kilome- within the ranges of basaltic meteorites from 77058, USA. 17Honeybee Robotics & Spacecraft Mechanisms Corporation, Pasadena, CA 91103, USA. 18Universität Bayreuth ter, these outcrops are similar in physical ap- Mars [Fe/Mn ratios, 36 to 45; Mg numbers, 24 und Eberhard Karls Universität, 72076 Tübingen, Germany. pearance and elemental chemistry; we interpret to 68 (fig. S1)] and indicate that any alteration of *To whom correspondence should be addressed. E-mail: them to represent the dominant surface rock type the protolith of these rocks did not substantially [email protected] of Cape York. mobilize Mg, Mn, or Fe. P contents are higher than

570 4 MAY 2012 VOL 336 SCIENCE www.sciencemag.org REPORTS those of basaltic sand, and Cr contents are lower. blocks that extends to the southeast. The ejecta are higher and quite variable. One location has However, these minor elements follow trends block Tisdale was investigated in detail. Tisdale the highest Zn concentration (~6300 mgg−1)of of martian basaltic and lherzolitic meteorites differs texturally and compositionally from all analyses from Mars. The Zn and Ni elemental (fig. S1), which suggests that igneous fraction- Chester Lake and Greeley Haven rocks. Because trends are similar to trends in hydrothermally ation established the major and minor element it was excavated from Odyssey crater, Tisdale may altered rocks around Home Plate in Gusev crater concentrations. represent a deeper unit within the Shoemaker for- (15). Variable P enrichment may result from We interpret all of these rocks to be breccias mation. Tisdale and other Odyssey ejecta blocks metasomatism of silicate materials by P-rich formed during the Endeavour impact. The com- are breccias, with poorly sorted, closely packed, hydrothermal solutions (16) or acidic solutions positional similarity to Meridiani basaltic sand is angular to rounded clasts up to several centi- dissolving P from source rocks and reprecip- consistent with the view that both materials meters in size (Fig. 3). Tisdale lacks the extensive itating it (17). It is unlikely, however, that the broadly reflect the composition of the surficial fine-grained matrix of Chester Lake and other high P content simply reflects an igneous com- crust in this region. rocks, and contains lithic fragments over a wide position different from that of the Chester Although we cannot assess the degree of range of grain sizes. Lake target material, as the Tisdale targets do shock metamorphism of the clasts or the presence The clasts in Tisdale exhibit spectral varia- not follow typical igneous fractionation trends, of glass in the matrix, we note that the texture of bility beyond that expected from discontinuous such as that defined by martian meteorites these rocks is similar to the typical texture of dust coatings, including positive and negative (fig. S1C). suevite breccias (13)commoninimpactsettings near-infrared spectral slopes and some 903-nm It is instructive to compare the rim deposits of on Earth and the Moon. For terrestrial suevites, absorptions. Small, localized spots in Tisdale Endeavour Crater to those of the comparably particle shape fabrics have been reported that are and nearby rocks exhibit 860- and 535-nm ab- sized Ries impact structure in Germany (18). At oriented radial to the impact point and conse- sorptions possibly consistent with a ferric phase Ries, thin deposits of surficial suevite (19), often quently have been related to the emplacement and/or minor hydrated Mg/Fe silicates. On the just a few meters thick, overlie the Bunte Breccia flow (14). The linear fabric in Chester Lake is basis of its textural and color properties, we in- that dominates the impactites of the rim (20). The oriented within ~5° of radial to the center of En- terpret Tisdale as a lithic breccia that is possibly Bunte Breccia is a poorly sorted polymict lithic deavour, although we cannot rule out the pos- polymict. breccia. It represents the continuous ballistic sibility that this texture is erosional rather than APXS measurements of Tisdale were made at ejecta deposit and is derived primarily from the 21

primary. three locations on a relatively dust-free vertical uppermost lithologies of the target ( ). Rounded on June 1, 2012 Near the southern end of Shoemaker Ridge, face (Table 1). The results are similar to Chester clasts are common. The surficial suevite that Opportunity encountered Odyssey crater (Fig. 1). Lake and the Greeley Haven outcrops; however, overlies it was deposited late in the impact pro- Odyssey is elliptical, ~23 m by 19 m, with its concentrations of Mg are lower, and concen- cess, either via fallout from a gaseous ejecta major axis oriented orthogonal to a field of ejecta trations of P and the trace elements Ni, Zn, and Br plume (22) or as a surface-hugging flow (23). We suggest that a similar relationship may hold at Endeavour, with Tisdale representing the main breccia unit of the rim, and Chester Lake and the rocks near Greeley Haven emplaced later in the impact flow. www.sciencemag.org Tisdale’s Zn abundances correlate with both S and P, which suggests that Zn sulfides, sulfates, and/or phosphates could be present. High-Zn materials on Earth are common in settings such as volcanogenic massive sulfide deposits, where hydrothermal circulation has mobilized Zn and caused precipitation of Zn sulfides and develop-

ment of associated alteration products (24). The Downloaded from heating caused by an impact the size of En- deavour is sufficient to cause hydrothermal cir- culation if water is present (25). We suggest that the Zn enrichment in Tisdale could have resulted from such activity. Enrichments of trace metal mineralization by hydrothermal fluids will nat- urally be localized where fluids readily flow, including fractures, zones with increased per- meability, and other void spaces. Highly hetero- geneous distribution of secondary mineralization is therefore expected. When normalized to the average composition of the sulfate-rich sandstones that constitute the Burns formation of the Meridiani plains (2, 3), all rocks of the Shoemaker formation show fraction- ations among the major elements—for example, high Na/Mg and Al/Mg ratios (fig. S2). Either the Shoemaker formation is not representative of the feedstock for the sulfate-rich grains of the Fig. 1. Opportunity’s traverse along the rim segment of Endeavour Crater named Cape York. Major features Burns formation, or the alteration process that discussed in the text are indicated. Image was acquired by the Mars Reconnaissance Orbiter HiRISE camera. formed those grains resulted in major composi- North is at the top. tional changes.

www.sciencemag.org SCIENCE VOL 336 4 MAY 2012 571 REPORTS

The composition of Shoemaker formation rocks is also not a good match to either Bounce Rock (26) or Marquette Island (27), two basaltic rocks encountered earlier in Opportunity’smis- sion that lie atop Burns formation sandstones. These must be ejecta blocks from distant craters that postdate Endeavour, sampling different crust- al materials. Cape York is encircled by a gently outward- sloping topographic bench ~6 m wide on the west and up to 20 m wide on the east (Fig. 1). The outer part of the bench on the western side ex- poses bright, thinly bedded sandstones with bedding that dips shallowly toward the plains. These sandstones lie directly above darker gran- ular sedimentary rocks that form the inner por- tion of the bench. This stratigraphic relationship is interpreted as the unconformable onlapping contact of the Burns formation onto older sedi- mentary materials shed from the Shoemaker for- mation. The inner bench materials, in turn, overlie the Noachian breccias that form the lower slopes of Cape York. Bench materials are cut in many places by bright linear veins. Veins are prominent in the

poorly exposed dark sedimentary materials of the on June 1, 2012 inner bench, but occur within the bright outcrops Fig. 2. Pancam false-color mosaic of the rock Chester Lake acquired on sol 2709 using the 753-, 535-, of the basal Burns sandstone as well. Measure- and 432-nm filters, sequence p2394. Scale across the image is ~1 m. Square shows location of MI ments of 37 veins yielded a mean width of 2 cm image (inset) acquired on sol 2713. Scale across MI image is ~3 cm. Clasts in Chester Lake, which stand and a mean exposed length of 33 cm. Most vein out in relief, are poorly sorted and irregularly shaped. The MI image shows irregular to cuspate fluting orientations lie subparallel to the margins of that is aligned with the orientations of elongate clasts. Cape York. Opportunity investigated one of these veins, named Homestake (Fig. 4), near the northern end of Cape York. Homestake forms a discontinuous, www.sciencemag.org flat-topped ridge 1 to 1.5 cm wide and ~50 cm long. It stands up to ~1 cm above the surrounding bedrock, which suggests that it is more erosion- resistant than the material into which it was em- placed. Microscopic Imager (MI) images reveal a fine linear texture perpendicular to the trend of the vein (Fig. 4B).

Three separate APXS analyses were con- Downloaded from ducted on Homestake. All were similar (Table 1), with high abundances of SO3 and CaO. The SO3/CaO ratio is within a few percent of stoichi- ometry for CaSO4, with a possible slight excess of sulfate. Other cations such as Mg, K, Al, and Fe do not show positive correlation with S, but Na does, so a small amount of Na sulfate could be present. The nonsalt portion of the APXS measure- ment of Homestake is dominated by Si and Al, which are highly correlated with each other (R2 = 0.89) and are anticorrelated with S (R2 =0.97to 0.99). MI images (Fig. 4B) show, however, that contaminants must be present in this measure- ment. Homestake does not fill the ~3.8-cm circular field of view of the APXS, so contamination by background material (mostly dark sand in the Fig. 3. Ejecta blocks from Odyssey crater. (A) Pancam false color image of the block Tisdale acquired on sol MI images) must be present. MI images of 2690 using the 753-, 535-, and 432-nm filters, sequence p2387. Height of Tisdale is ~30 cm. Square Homestake show a mottled upper surface, also shows location of image in (B). (B) MI image acquired on sol 2696. Scale across the image is ~2.5 cm. (C) suggesting surface contamination. Pancam false-color image of the ejecta block Kidd Creek acquired on sol 2694 using the 753-, 535-, and Modeling the composition of Homestake as 432-nm filters, sequence p2593. Scale across Kidd Creek is ~1 m. All images show a lithic breccia texture, a three-component mixture of calcium sulfate, with clasts ranging in size from tens of centimeters to near the resolution limit of MI images.

572 4 MAY 2012 VOL 336 SCIENCE www.sciencemag.org REPORTS typical Meridiani basaltic sand, and typical martian 43% CaSO4, 29% sand, and 28% dust 4 mmthick itself is predominantly CaSO4, with perhaps minor dust, and accounting for the energy dependence of (28). Because the MI images are consistent with Na2SO4, phosphate, and a Cl-containing salt. the signal from a thin dust layer, we find that the these amounts of background sand and surface dust Calcium sulfates can have a range of hydration composition in Table 1 is matched by a mixture of contamination, we conclude that the vein material states. APXS data do not constrain the hydration

Table 1. Elemental chemistry as determined by the APXS instrument, under similar rocks near Greeley Haven. Tisdale is a lithic breccia with variable the standard assumption of a homogeneous matrix. Chester Lake is a suevite- composition. Homestake is the vein in Fig. 4, and Deadwood is the material in like breccia near Spirit Point (Fig. 1), and Transvaal and Boesmanskop are which Homestake is emplaced.

Chester Lake Transvaal Boesmanskop Basaltic Effective Komati Tisdale‡ Homestake§ Deadwood Dust# Matrix* Clast (matrix) (matrix)† (clast) sand|| dust** Weight percent

Na2O2.7T 0.2 2.7 T 0.2 2.3 T 0.3 2.4 T 0.4 2.4 T 0.2 1.8 T 0.3 1.6 T 0.2 2.2 T 0.2 2.34 2.31 2.28 2.2 T 0.3 2.5 T 0.4 MgO 8.8 T 0.1 7.4 T 0.1 7.5 T 0.2 8.9 T 0.2 8.4 T 0.1 6.2 T 0.1 4.7 T 0.1 5.7 T 0.1 7.33 7.62 7.18 6.0 T 0.1 6.2 T 0.2

Al2O3 8.8 T 0.1 10.1 T 0.2 9.2 T 0.2 9.5 T 0.2 9.3 T 0.2 8.9 T 0.2 4.8 T 0.1 8.3 T 0.2 9.65 9.21 7.99 10.0 T 0.2 10.1 T 0.2

SiO2 45.5 T 0.4 46.1 T 0.5 45.9 T 0.5 45.6 T 0.5 44.7 T 0.4 42.6 T 0.5 25.3 T 0.3 44.0 T 0.5 46.97 45.33 35.29 46.2 T 0.5 45.4 T 0.5

P2O5 1.00 T 0.08 1.13 T 0.09 1.04 T 0.10 1.21 T 0.09 1.16 T 0.09 3.14 T 0.13 0.78 T 0.11 1.13 T 0.10 0.85 0.91 0.73 1.22 T 0.10 on June 1, 2012 1.20 T 0.10

SO3 3.1 T 0.1 4.8 T 0.1 6.3 T 0.1 5.6 T 0.1 5.8 T 0.1 8.6 T 0.2 32.7 T 0.4 9.2 T 0.1 4.68 7.23 5.02 6.0 T 0.1 6.5 T 0.1 Cl 0.9 T 0.0 1.0 T 0.0 1.0 T 0.0 1.0 T 0.0 0.9 T 0.0 1.2 T 0.0 1.0 T 0.0 1.1 T 0.0 0.59 0.81 0.48 0.9 T 0.0 1.0 T 0.0

K2O 0.41 T 0.06 0.46 T 0.06 0.49 T 0.06 0.51 T 0.06 0.58 T 0.06 0.43 T 0.07 0.27 T 0.06 0.62 T 0.07 0.51 0.5 0.19 0.50 T 0.07 www.sciencemag.org 0.53 T 0.07 CaO 6.8 T 0.1 7.1 T 0.1 6.6 T 0.1 5.7 T 0.1 6.1 T 0.1 7.1 T 0.1 22.0 T 0.2 6.7 T 0.1 7.38 6.66 2.03 6.8 T 0.1 5.9 T 0.1

TiO2 1.09 T 0.08 1.15 T 0.08 1.11 T 0.08 1.03 T 0.08 1.08 T 0.07 0.99 T 0.08 0.27 T 0.06 0.98 T 0.08 0.9 0.99 0.22 1.05 T 0.10 1.05 T 0.09 Downloaded from Cr2O3 0.25 T 0.03 0.27 T 0.04 0.28 T 0.04 0.25 T 0.04 0.18 T 0.03 0.16 T 0.04 0.13 T 0.04 0.30 T 0.04 0.39 0.34 0.05 0.27 T 0.04 0.23 T 0.04 MnO 0.48 T 0.02 0.63 T 0.02 0.41 T 0.02 0.41 T 0.02 0.59 T 0.01 0.38 T 0.02 0.15 T 0.01 0.22 T 0.02 0.39 0.36 0.04 0.38 T 0.02 0.23 T 0.02 FeO 20.1 T 0.1 17.0 T 0.2 17.8 T 0.2 17.6 T 0.2 18.7 T 0.1 17.6 T 0.2 6.4 T 0.1 19.4 T 0.2 17.57 17.59 1.62 18.0 T 0.2 18.8 T 0.2 Concentration (µg g–1) Ni 482 T 54 461 T 65 565 T 64 615 T 65 461 T 48 950 T 84 36 T 83 410 T 62 349 486 53 1405 T 95 2030 T 108 Zn 246 T 18 244 T 23 294 T 23 350 T 25 266 T 15 6267 T 108 121 T 20 521 T 29 199 390 29 1798 T 58 710 T 42 Br 124 T 18 68 T 19 139 T 21 153 T 22 229 T 19 779 T 35 68 T 19 301 T 25 24 31 1 722 T 34 377 T 29 *Sample abraded with the RAT. †Sample brushed with the RAT. ‡Measurements at three different locations on a relatively dust-free vertical face. §Average of three measurements. ||Typical Meridiani basaltic sand composition. #Typical martian dust used in geochemical modeling of Homestake. **Effective composition of a layer of dust 4 mm thick, taking into account the energy dependence of the APXS signal.

www.sciencemag.org SCIENCE VOL 336 4 MAY 2012 573 REPORTS

state of Homestake because of the instrument’s unsaturated Pancam observations acquired at dif- drite lacks a hydration band, and the weak bas- inability to detect H and the confounding ef- ferent viewing geometries and incidence angles, sanite hydration band is centered near ~950 nm, fects of background and contamination discussed convincing us that the signature is due to hydra- between Pancam’s two longest-wavelength band above. However, Pancam’s longest-wavelength tion in Homestake rather than any instrumental, centers. filter (1009 T 19 nm) provides the ability to detect calibration, or viewing geometry artifact (30). The Gypsum veins have been reported in a variety remotely and map spatially certain hydrated min- magnitude of the hydration signature and the al- of settings on Earth, where their formation is erals, based on the presence of the 2n1 + n3 H2O bedo both increased after Opportunity drove over invariably attributed to precipitation from rela- combination absorption band and/or the 3vOH and exposed fresh crushed parts of Homestake; tively low-temperature (less than ~60°C) water in overtone absorption band centered near ~1000 nm this suggests that the hydrated material is not a fractures (31–34). Vein growth is often antitaxial, in many minerals containing bound H2Oand/or superficial coating or rind, but rather a component with nucleation occurring at the vein-wall inter- OH – (29, 30). of the bulk volume of the vein material. face. Crystals precipitated in such settings are Figure 5A shows that hydrated mineral sig- Laboratory reflectance spectra of calcium sul- commonly fibrous, with long axes that track the natures [characterized by four total spectral pa- fates convolved to Pancam bandpasses (Fig. 5B) opening trajectory of the fracture (34–36); fibers rameters (30), including steeply negative slopes suggest that the hydration signature is consistent therefore form perpendicular to the vein axis when from 934 to 1009 nm] are detected in Homestake. with gypsum (CaSO4•2H2O), but not anhydrite vein growth occurs in extension fractures. We These spectral features occur in three separate (CaSO4) or bassanite (CaSO4•0.5H2O). Anhy- suggest that the transverse lineations seen in MI images of Homestake are remnants of such a texture. The orientations of the veins themselves Fig. 4. (A)Pancamapprox- suggest that the rocks of the bench surrounding imate true-color image of the vein Homestake acquired Cape York were subjected to horizontal tension on sol 2769 using the 753-, perpendicular to the bench margins at the time of 535-, and 432-nm filters, se- vein emplacement, perhaps related to sediment quence p2574. Scale across compaction, dewatering, and settling. the image is ~40 cm. Rectan- Homestake was emplaced in the darker inner gle shows location of image unit of the bench surrounding Cape York. The

in (B). (B)MImosaicacquired rock of this unit is platy in appearance, with on June 1, 2012 on sol 2766. millimeter-scale layering that is poorly exposed but shows locally varying strikes and dips. Its elemental composition, measured at a location named Deadwood (Table 1), is similar to that of rocks of the Shoemaker formation, although without strong Zn and Ni enrichments. From its elevated Ca and S concentrations, Deadwood also appears to contain a small amount (~10%) of Homestake-like material. Typical Meridiani ba- www.sciencemag.org saltic sand could constitute up to 30% of the total (MI images suggest 20 to 30% contamination), but not more because the Mn/Fe ratio would become untenably low. We interpret Deadwood to be a clastic sedimentary rock dominated by grains from the Shoemaker formation, with minor Downloaded from Fig. 5. (A)Pancam“hydration signature” data overlain on a 754-nm image of Homestake acquired on sol 2769, sequence p2574. Colors indicate regions where the 934- to 1009-nm spectral slope is negative and where four other hydration sig- nature spectral parameters (29) are also met. (B)Comparisonof Pancam relative reflectance [R*(41)] spectra of Homestake to laboratory reflectance spectra (42) of three calcium sulfates with different levels of hydration. Solid lines are full-resolution lab spectra; points indicate lab spectra values convolved to Pancam bandpasses. The anhydrite, gypsum, and bassanite data are offset by +0.2, 0.0, and –0.2 reflectance units, respectively. The inset shows the magnitude of the negative 934- to 1009-nm spectral slope values for these three mineral samples compared to the minimum, average, and maximum Pancam Homestake values. Error bars represent estimated 1s uncertainties on the laboratory and Pancam ratio values. Logarithmic plot; units are 10−4 nm−1.

574 4 MAY 2012 VOL 336 SCIENCE www.sciencemag.org REPORTS

CaSO4, perhaps as cement. Basaltic sand con- of gypsum-anhydrite transformations further References and Notes taminates the APXS measurement but does not suggests that these rocks remained at relatively 1. The Noachian is a geologic system on Mars representing the oldest period of the planet’s geologic history, change this interpretation. low temperatures since vein emplacement. characterized by high impact rates. The key stratigraphic and crosscutting rela- Development of the materials investigated 2. S. W. Squyres et al., J. Geophys. Res. 111, E12S12 tionships along the western margin of Cape York at Endeavour began with an impact into ba- (2006). are depicted in Fig. 6. We interpret the inner saltic rocks, producing breccias with a stratigra- 3. R. E. Arvidson et al., J. Geophys. Res. 116, E00F15 portion of the bench, characterized by Deadwood, phy similar to that observed at some comparably (2011). 4. J. J. Wray et al., Geophys. Res. Lett. 36, L21201 to be the first sedimentary unit to form on Endeav- sized terrestrial impact craters. Degradation and (2009). our’s rim, dominated by material shed from the shedding of these breccias formed thin sedimen- 5. A sol is defined as one martian solar day. Cape York breccias. This material, and the basal tary materials immediately surrounding Cape 6. Shoemaker Ridge is named to honor the late Burns formation sandstone that unconformably York, over which the basal Burns formation Eugene M. Shoemaker, one of the founders of planetary geoscience. overlies it, were subsequently cut by fractures that sandstones were unconformably deposited. Em- 7. Greeley Haven is named to honor the late Ronald were filled by gypsum precipitated from sulfate- placement of the gypsum veins then took place, Greeley, distinguished planetary scientist and member rich fluids generated within the nearby Noachian postdating the earliest Burns formation sand- of the Athena Science Team. crust. These fluids were likely at a low temper- stones but probably predating much of the rest 8. S. W. Squyres et al., J. Geophys. Res. 108, 8062 (2003). ature (if hydrothermal, they were distal), because of the Burns formation stratigraphy encountered 9. The Miniature Thermal Emission Spectrometer anhydrite would be expected otherwise. In pure by Opportunity. (Mini-TES) was not used for this investigation because water the gypsum/anhydrite conversion temper- The ubiquity of impact breccia at Cape York of temperature-related instrument degradation and ature is ~40° to 60°C, and in concentrated brines contrasts with the only other Noachian terrain optically thick dust on its mirrors accumulated during a it can be substantially lower (37). Calcium sul- explored in situ, the Columbia Hills in Gusev prior global dust storm. The Mössbauer Spectrometer was not used because of the decay of its 57Co fate was precipitated closest to the Noachian Crater. The rover Spirit encountered great litho- radiation source. source rocks, rather than other sulfates (e.g., logic diversity there, including materials inter- 10. T. Myrick et al., in Proceedings of AIAA Space 2004 MgSO •nH O; FeSO •nH O) or chlorides, be- preted as impact ejecta (11). However, none were Conference and Exhibit (American Institute of Aeronautics 4 2 4 2 – cause of its lower solubility. Some gypsum was breccias, and none had the lateral extent of the and Astronautics, Reston, VA, 2004), pp. 1 11. 11. S. W. Squyres et al., J. Geophys. Res. 111, E02S11 also precipitated as cement in the bedrock. Un- Shoemaker formation. We suggest that the differ- (2006). ’

like the Burns formation sulfates that dominate ence can be attributed to Opportunity s sampling 12. R. E. Arvidson et al., Science 306, 1730 (2004). on June 1, 2012 the Meridiani plains and are rich in jarosite (2, 3), of the rim deposits of a single large crater, rather 13. D. Stöffler, R. A. F. Grieve, in Metamorphic Rocks: the gypsum of Homestake did not require acidic than Spirit’s sampling of more distal ejecta from A Classification and Glossary of Terms, Recommendations of the International Union of Geological Sciences, fluids for its formation. We suggest, however, multiple impacts. D. Fettes, J. Desmons, Eds. (Cambridge Univ. Press, that the fluids from which the gypsum was pre- The gypsum veins at Cape York provide clear Cambridge, 2007), chap. 2.11. cipitated may have been a contributor to the evidence for relatively dilute [water activity aw ≈ 14. C. Meyer, M. Jebrak, D. Stoffler, U. Riller, Geol. Soc. overall hydrologic budget responsible for for- 0.98 (39)] water at a moderate temperature, per- Am. Bull. 123, 2312 (2011). 15. M. E. Schmidt et al., J. Geophys. Res. 113, E06S12 mation of the Meridiani sulfate sandstones. haps supporting locally and transiently habitable (2008). The absence of substantial deformation of the environments. More broadly, rocks at Cape York 16. C. G. Wheat, R. A. Feely, M. J. Motti, Geochim. veins suggests that there has been minimal trans- appear to record early events in a transition from Cosmochim. Acta 60, 3593 (1996). www.sciencemag.org formation between gypsum and anhydrite since (commonly) hydrothermal waters that altered 17. E. S. Deevey Jr., Sci. Am. 223, 148 (1970). emplacement. The molar volumes of these min- basaltic crust to phyllosilicates (40)tosulfate- 18. J. Pohl et al., in Impact and Explosion Cratering, D. J. Roddy, R. O. Pepin, R. B. Merrill, Eds. (Pergamon, erals differ by a factor of ~1.6; in terrestrial charged ground waters that generated salt-rich New York, 1977), pp. 343–404. settings, such transformations typically produce sandstones deposited widely over the Meridiani 19. W. von Engelhardt et al., Meteoritics 30, 279 complex deformation features (38). The absence plains and elsewhere. (1995). 20. W. von Engelhardt, Tectonophysics 171, 250 (1990). 21. F. Hörz, R. Ostertag, D. A. Rainey, Rev. Geophys. 21, 1667 (1983).

22. W. von Engelhardt, Meteorit. Planet. Sci. 32, 545 Downloaded from (1997). 23. H. E. Newsom, G. Graup, T. Sewards, K. Keil, J. Geophys. Res. 91, E239 (1986). 24. J. M. Franklin, J. W. Lydon, D. F. Sangster, Econ. Geol. (75th anniv. volume), 485 (1981). 25. J. A. Rathbun, S. W. Squyres, Icarus 157, 362 (2002). 26. J. Zipfel et al., Meteorit. Planet. Sci. 46 (suppl. S1), A1 (2011). 27. D. W. Mittlefehldt et al., Lunar Planet. Sci. 41, 2109 (2009). 28. A 4-mm-thick dust layer is the simplest model that fits the APXS data; the actual dust coating is probably discontinuous and variable in thickness. 29. A. Wang et al., J. Geophys. Res. 113, E12S40 (2008). 30. M. S. Rice et al., Icarus 205, 375 (2010). 31. S. Taber, J. Geol. 26, 56 (1918). 32. T. C. Gustavson et al., J. Sed. Res. A 64,88 (1994). Fig. 6. Schematic east-west cross section, with large vertical exaggeration, through the western flank of 33. M. El Tabakh et al., J. Sed. Res. 68, 88 (1998). Cape York. Materials shed from the Shoemaker formation breccias form the lowest sedimentary unit, 34. S. L. Philipp, Geol. Mag. 145, 831 (2008). which includes Deadwood. This unit is overlain by the sulfate-rich sandstones of the Burns formation; 35. D. W. Durney, J. G. Ramsay, in Gravity and Tectonics,K.A.deJong,R.Scholten,Eds.(Wiley, the heavy line separating them represents an unconformity. The Deadwood unit and the basal portion New York, 1973), pp. 67–96. of the Burns formation closest to Cape York are cut by fractures; these fractures are filled with gypsum 36.P.D.Bons,M.Montenari,J. Struct. Geol. 27, 231 that was precipitated from waters arising from the underlying breccias. (2005).

www.sciencemag.org SCIENCE VOL 336 4 MAY 2012 575 REPORTS

37. D. Freyer, W. Voigt, Monatsh. Chem. 134, 693 (2003). 42. R. N. Clark et al., U.S. Geological Survey, USGS Digital Supplementary Materials 38. W. M. Bundy, J. Sed. Res 26, 240 (1956). Spectral Library 06 (2007); http://speclab.cr.usgs. www.sciencemag.org/cgi/content/full/336/6081/570/DC1 39. N. J. Tosca, A. H. Knoll, S. M. McLennan, Science 320, gov/spectral.lib06. Figs. S1 and S2 1204 (2008). References (43–96) 40. B. L. Ehlmann et al., Nature 479, 53 (2011). Acknowledgments: This research was carried out for the 41. J. F. Bell III et al., J. Geophys. Res. 108, 8063 Jet Propulsion Laboratory, California Institute of Technology, 13 February 2012; accepted 3 April 2012 (2003). under a contract with NASA. 10.1126/science.1220476

with linear regressions for all available data for 21st-Century Evolution of Greenland 2000 to 2005 (111 glaciers) and 2005 to 2010 (123 glaciers) to evaluate trends and fill data gaps Outlet Glacier Velocities (figs. S2 and S3). Our record reveals the complexity of Green- land’s ice flow. Greenland’s largest land-terminating T. Moon,1,2* I. Joughin,2 B. Smith,2 I. Howat3,4 glaciers are located primarily along the south- west coast, with a few in the northeast (Fig. 1). Earlier observations on several of Greenland’s outlet glaciers, starting near the turn of the 21st Nearly all flow at peak velocities between 10 century, indicated rapid (annual-scale) and large (>100%) increases in glacier velocity. Combining and 100 m/year, so that annual changes of 10 data from several satellites, we produce a decade-long (2000 to 2010) record documenting the to 30 m/year produce large long-term trends ongoing velocity evolution of nearly all (200+) of Greenland’s major outlet glaciers, revealing (>15% change over 5 years). Most (70%) of the complex spatial and temporal patterns. Changes on fast-flow marine-terminating glaciers contrast land-terminating glaciers with a notable trend with steady velocities on ice-shelf–terminating glaciers and slow speeds on land-terminating slowed during 2005 to 2010—a continuing trend glaciers. Regionally, glaciers in the northwest accelerated steadily, with more variability in the for half of them. The scale of these changes, southeast and relatively steady flow elsewhere. Intraregional variability shows a complex response however, is close to the measurement error and to regional and local forcing. Observed acceleration indicates that sea level rise from Greenland seasonal variability (11) and orders of magnitude

may fall well below proposed upper bounds. smaller than changes seen on many fast-flowing on June 1, 2012 glaciers. The one outlier, Frederikshab Isblink hanges in glacier dynamics contribute to sheet–wide scale (13, 14) or smaller regions with (fig. S1), has a large lobe-shaped terminus that is roughly half of the Greenland Ice Sheet’s more frequent sampling (7). Where comprehen- primarily land-terminating, but with one segment Ccurrent mass loss (~250 Gt/year, equiv- sive records exist, they have been used to focus of lake-terminating ice front. The velocity field alent to 0.6 mm/year sea level rise) (1, 2), largely on aggregate discharge rather than regional var- suggests that this segment helps the glacier through thinning and increased calving as gla- iability (2). We present a decade-long record, maintain a higher peak velocity (~270 m/year) ciers have sped up. Large changes in ice dy- with annual sampling for the latter half, to ex- than other land-terminating glaciers and hints namics have been observed (3), but were not amine decadal-scale trends and regional and local at the importance of a calving terminus in main- accounted for in early models and led to the interannual variation, and to inform predictions taining fast ice flow. www.sciencemag.org inability to quantify uncertainty of 21st-century of sea level rise. Ice-shelf–terminating glaciers (Fig. 1) have sea level rise in the Intergovernmental Panel on To create this record, we produced velocity mean velocities (300 to 1670 m/year) that are Climate Change’s (IPCC’s) Fourth Assessment maps for winter 2000 to 2001 (referred to as generally slower than those of other marine- (4). Although multiglacier speedups have been 2000) and annually for each winter from 2005 terminating glaciers (total mean: 1890 m/year), linked to recent warming in Greenland (5–7), to 2006 through 2010 to 2011 (referred to using but most show negligible change for 2000 to the exact connection to climate change is poorly the earlier year for the map), using synthetic 2010. The most notable change occurred on known, but may be related to processes con- aperture radar data from the Canadian Space Hagen Bræ (from 50 m/year in 2000 to 650

trolled by ice-ocean interaction (8–10). A firm Agency’s RADARSAT-1, German TerraSAR-X, m/year in 2007), a previously identified surge- Downloaded from understanding of the processes driving recent and Japanese Advanced Land Observation Sat- type glacier (18). change, which is needed to improve predic- ellite (ALOS) (table S1). We used a combination Surge-type glaciers occur mostly in the north- tions of sea level rise, requires a better charac- of speckle-tracking and interferometric algorithms west, north, and east (18, 19). In several cases, terization of the temporal and spatial patterns to estimate ice-flow velocity (14, 15). Coverage 1- or 2-year velocity changes suggest surge- of ice flow across the ice sheet. for each year is almost complete, with some un- type behavior, as observed on Harald Moltke Despite the need for comprehensive data, avoidable gaps, primarily in the south, due to Bræ (high speed in 2005), where surges have recent studies of glacier velocity in Greenland satellite acquisition limits. been recorded before, and Adolf Hoel Gletscher are limited in spatial and temporal resolution. Of the 206 largest Greenland outlet glaciers, (low speed in 2007) (fig. S1) and Kangilerngata For Jakobshavn Isbræ, Helheim Gletscher, and 178 have adequate temporal coverage for 2000 sermia (low speed in 2005), where earlier surges Kangerdlugssuaq Gletscher, three of Greenland’s to 2005, and 195 have sufficient data for 2005 to have not been recorded. Other glaciers where fastest outlet glaciers, velocity is relatively well 2010 (Fig. 1) (16). We divided these glaciers surges have been observed previously, includ- documented (3, 11, 12).FormostofGreenland’s into several categories. First, we identified land- ing Storstrommen and L. Bistrup Bræ (18) and other 200+ outlet glaciers, however, observation terminating, ice-shelf–terminating (ice shelf >10 km Sortebræ (20), maintained quiescent speeds over has been limited to ~5-year sampling on an ice- long), and low-velocity marine-terminating (mean the past decade. velocity <200 m/year) glaciers (55 total). Next, Most glaciers in east Greenland are marine- 1Earth and Space Sciences, University of Washington, Seattle, glaciers with highly variable behavior were terminating, but have substantially slower mean WA 98195, USA. 2Polar Science Center, Applied Physics Lab, separated to avoid misrepresenting large varia- velocities (1040 m/year) relative to southeast 3 University of Washington, Seattle, WA 98105, USA. School of tions as consistent trends (16). This included (2830 m/year) or northwest (1630 m/year) marine- Earth Sciences, Ohio State University, Columbus, OH 43210, USA. 4Byrd Polar Research Center, Ohio State University, Co- glaciers such as Harald Moltke Bræ (fig. S1), terminating glaciers. This is consistent with the lumbus, OH 43210, USA. where apparent surge behavior produces erratic lower regional discharge from this low accu- *To whom correspondence should be addressed. E-mail: changes (17). The final group consisted of fast- mulation area (21). As a group, eastern glaciers [email protected] flow marine-terminating glaciers that we fit showed only negligible changes from 2000 to

576 4 MAY 2012 VOL 336 SCIENCE www.sciencemag.org